|
|
|
@ -159,10 +159,19 @@ Enhancements
|
|
|
|
|
``mitogen_get_stack`` action that is automatically added to the action
|
|
|
|
|
plug-in path. See :ref:`mitogen-get-stack` for more information.
|
|
|
|
|
|
|
|
|
|
* Multiplexer threads are pinned to one CPU, reducing latency and SMP overhead
|
|
|
|
|
on a hot path exercised for every task. This yielded an 8% speedup in a
|
|
|
|
|
64-target job composed of many short tasks, and should easily be visible as a
|
|
|
|
|
runtime improvement in many-host runs.
|
|
|
|
|
* `152effc2 <https://github.com/dw/mitogen/commit/152effc2>`_: multiplexer
|
|
|
|
|
threads are pinned to one CPU, reducing latency and SMP overhead on a hot
|
|
|
|
|
path exercised for every task. This yielded an 8% speedup in a 64-target job
|
|
|
|
|
composed of many short tasks, and should easily be visible as a runtime
|
|
|
|
|
improvement in many-host runs.
|
|
|
|
|
|
|
|
|
|
* `#260 <https://github.com/dw/mitogen/issues/260>`_,
|
|
|
|
|
`a18a083c <https://github.com/dw/mitogen/commit/a18a083c>`_: brokers no
|
|
|
|
|
longer wait for indication of readiness from the OS to transmit data, and
|
|
|
|
|
instead assume transmission will probably succeed. As this is often true, one
|
|
|
|
|
IO loop iteration and two poller reconfigurations are avoided for many
|
|
|
|
|
messages, yielding a significant reduction in interprocess round-trip
|
|
|
|
|
latency.
|
|
|
|
|
|
|
|
|
|
* `#415 <https://github.com/dw/mitogen/issues/415>`_,
|
|
|
|
|
`#493 <https://github.com/dw/mitogen/issues/493>`_: the interface employed
|
|
|
|
|